A method for automatically evaluating multi-attributes of seismic acquisition data
Patent Information
- Application Number
- CN202510495661.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-21
- Publication Date
- 2026-10-23
AI Technical Summary
[0003]传统地震数据质控方法多依赖人工经验，针对单一属性(如信噪比、分辨率)设置固定阈值进行评价，存在以下显著缺陷：首先，人工评分易受个体经验差异影响，尤其在复杂工区(如强噪声环境或多断层区域)，不同专家对同一数据的评价可能产生分歧，导致质控结果不稳定
[0060]本发明的有益效果为：通过计算多个属性对评分的贡献度和属性之间的伴随度筛选评价属性，主观性因素影响小，模型收敛速度快且准确性高。使用神经网络建立多属性评价模型，回避了现有技术对单一属性阈值主观设置存在的争议。通过数学统计方法筛选出能够对地震资料品质贡献度高的属性，并去除与其他属性存在相互影响的属性，提高了多属性评价的科学性。通过人工评级结合客观评分结果获取综合评分，实现了基于多属性的量化评价。通过计算方差膨胀因子量化属性间的线性相关性，筛选出不存在相互影响的评价属性，确保评价属性彼此独立，避免模型陷入局部最优解，产生预测偏差。
Smart Images
Figure CN122962639A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of seismic data acquisition and processing technology for seismic exploration in oil and gas fields, and in particular to an automatic multi-attribute evaluation method for seismic acquisition data. Background Technology
[0002] Seismic data acquisition is a fundamental step in fields such as oil and gas exploration, geological surveys, and earthquake monitoring, and the quality of the data directly affects the reliability of subsequent processing and interpretation.
[0003] Traditional seismic data quality control methods largely rely on human experience, setting fixed thresholds for single attributes (such as signal-to-noise ratio and resolution) for evaluation. This approach has several significant drawbacks: First, human scoring is easily influenced by individual experience differences, especially in complex work areas (such as high-noise environments or multi-fault zones). Different experts may disagree on the same data, leading to unstable quality control results. Second, a single attribute (such as signal-to-noise ratio) cannot comprehensively reflect data quality. For example, high signal-to-noise ratio data may fail to identify thin geological structures due to insufficient resolution, and current technologies lack multi-attribute collaborative evaluation mechanisms. Finally, in multi-attribute evaluations, strong correlations exist between some attributes (such as shot energy and acoustic energy), leading to model overfitting or distorted contribution allocation, reducing evaluation reliability.
[0004] How to solve the above-mentioned technical problems is the challenge facing this invention. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides an automatic multi-attribute evaluation method for seismic acquisition data. This method filters evaluation attributes by calculating the contribution of multiple attributes to the score and the degree of association between attributes, minimizing the influence of subjective factors, and achieving fast model convergence and high accuracy.
[0006] The technical solution adopted by this invention to solve its technical problem is as follows: This invention provides an automatic multi-attribute evaluation method for seismic acquisition data, comprising the following steps:
[0007] Collect and preprocess earthquake data to establish a sample earthquake database;
[0008] The sample seismic data is scored by selecting characteristic attributes of the sample seismic data;
[0009] The feature attributes are divided into positive and negative attributes according to their impact on the seismic data scoring, and then normalized.
[0010] Calculate the contribution of feature attributes to the overall score and obtain sensitive attributes;
[0011] Calculate the degree of association between sensitive attributes, analyze whether there is mutual influence between attributes, screen out evaluation attributes, and build and train a multi-attribute comprehensive evaluation model based on the evaluation attributes.
[0012] Input the normalized evaluation attribute values and use a multi-attribute comprehensive evaluation model to score them.
[0013] Preferably, the preprocessing includes outlier removal, noise reduction, and standardization; the seismic data in the sample seismic database covers more than 95% of the recorded features within the work area, including seismic wave correlation features, temporal features, spatial features, and acquisition condition-related features.
[0014] It should be noted that seismic wave characteristics are the most fundamental seismic data, including the propagation mode, amplitude, frequency, and velocity of seismic waves. Temporal characteristics mainly involve information such as the time of occurrence of the seismic event, the energy released from the hypocenter, the location of the hypocenter, and the time interval between other seismic events. Spatial characteristics refer to the spatial distribution of seismic activity, including the epicenter location, focal depth, and fault distribution. Acquisition condition-related characteristics refer to factors related to the data acquisition process, including the accuracy of the acquisition equipment, environmental conditions (such as temperature and humidity), and sensor performance.
[0015] Outlier data refers to values that are significantly inconsistent with the majority of the data. For example, in seismic data, outliers may be caused by equipment malfunctions, acquisition errors, external interference, or other reasons. Statistical methods, such as the standard deviation method or box plot method, can be used to identify and remove data points that fall outside the normal range, thereby reducing noise and improving data quality.
[0016] Denoising processes eliminate irrelevant or interfering components in seismic data, making the data more accurate and thus improving the accuracy of subsequent analyses.
[0017] Z-score standardization or min-max standardization is used to unify data of different dimensions and scales, preventing some data from having a disproportionate impact on subsequent scoring due to their large scale, and avoiding the impact of scale differences on the final model or scoring results.
[0018] Preferably, the characteristic attributes include resolution, gun energy, signal-to-noise ratio, acoustic energy, surface wave energy, industrial interference channel, non-operating channel, and main frequency;
[0019] The specific steps for scoring the sample seismic data are as follows:
[0020] The sample seismic data were manually rated into four levels: unqualified, qualified, good, and excellent.
[0021] Select the time window where the main target layer is located, calculate the signal-to-noise ratio and resolution of the sample seismic data, normalize them respectively, and take the average value as the objective scoring result;
[0022] Based on the manual rating results of the sample earthquake data, the objective scoring results are segmented and normalized.
[0023] The sample earthquake data are calculated into a comprehensive score based on both manual rating results and objective scoring results, as shown in the following formula:
[0024]
[0025] In the formula, a i The value is the segmented normalized value of the objective score result for sample i.
[0026] It should be noted that segmented normalization of objective scoring results refers to mapping the continuous values of objective scores to specific intervals corresponding to the manual grading based on the four manually divided levels (unqualified, qualified, good, and excellent), so that the automatic scoring is aligned with the manual evaluation standards and the distribution differences of automatic scores within different levels are eliminated.
[0027] By combining manual rating with objective scoring results to obtain a comprehensive score, a quantitative evaluation based on multiple attributes is achieved.
[0028] Preferably, the positive attribute is positively correlated with the seismic data score, and the negative attribute is negatively correlated with the seismic data score;
[0029] The positive attribute normalization process uses the maximum value normalization method, and the formula is expressed as follows:
[0030]
[0031] In the formula, x i For the i-th positive attribute value, Let x be the i-th normalized positive attribute value. max This is the maximum value of the attribute.
[0032] The negative attribute normalization process is as follows: Negative attribute values are sorted in ascending order, the average of the bottom 10% of attribute values is selected as the standard value of the negative attribute, and the negative of the standard value is used for normalization. The formula is as follows:
[0033]
[0034] In the formula, y norm The standard value for negative attributes, y i Let N be the i-th negative attribute value, and N be the total number of samples of the last 10% of attribute values after the negative attribute values are sorted in ascending order.
[0035]
[0036] In the formula, Let be the i-th normalized negative attribute value.
[0037] It should be noted that by distinguishing between positive and negative attributes and normalizing them separately, all attributes are defined and evaluated on the same scale, reducing the order-of-magnitude difference between the values of each evaluation attribute, which helps with model fitting.
[0038] Preferably, the calculation of the contribution of the feature attributes to the comprehensive score, and the acquisition of sensitive attributes, specifically involves:
[0039] Grey relational analysis was performed on the characteristic attributes and the comprehensive score, and the characteristic attributes with a degree of 0.7 or greater were selected as sensitive attributes.
[0040] It should be noted that by using grey relational analysis to exclude weakly correlated attributes (feature attributes with a correlation degree of less than 0.7 with the comprehensive score), the model input dimensions are reduced, and key indicators are focused on, effectively improving the model training efficiency.
[0041] Preferably, the calculation of the degree of association between sensitive attributes, the analysis of whether there is mutual influence between attributes, and the selection of evaluation attributes specifically involve:
[0042] The degree of association between each sensitive attribute is obtained by calculating the variance inflation factor (VIF).
[0043]
[0044] In the formula, VIF i Let be the variance inflation factor for the i-th evaluation attribute, where i is the number of evaluation attributes. The correlation coefficient between the i-th evaluation attribute and the other k-1 evaluation attributes is used to treat the i-th evaluation attribute as the explained attribute.
[0045]
[0046] In the formula, The average value of the attribute being explained. Let z be the estimated value of the i-th explained attribute. i Let be the attribute value of the i-th interpreted attribute.
[0047] The judgment criteria are as follows:
[0048] When 1≤VIF≤5, there is no mutual influence between attributes;
[0049] When 5 < VIF ≤ 10, there is a weak mutual influence between attributes;
[0050] When 10 < VIF ≤ 100, there is a strong mutual influence between attributes;
[0051] When VIF > 100, there are serious mutual influences between attributes;
[0052] Sensitive attributes that do not have mutual influence are retained as evaluation attributes.
[0053] It should be noted that mutual interference between attributes leads to large fluctuations in model coefficients, making it difficult to distinguish the true contribution of each attribute and resulting in distorted evaluation results (e.g., the contribution of a certain attribute is incorrectly overestimated or underestimated). By calculating the variance inflation factor, the linear correlation between attributes is quantified, and evaluation attributes that do not influence each other are screened out. This ensures that the evaluation attributes are independent of each other, avoids the model getting trapped in local optima, and prevents prediction bias.
[0054] Preferably, the construction and training of the multi-attribute comprehensive evaluation model based on evaluation attributes specifically involves:
[0055] Using the evaluation attributes selected from the sample earthquake data as input and the comprehensive score as output, the original sample dataset is divided into n subsets and trained into a neural network through n iterations.
[0056] For the i-th iteration, the i-th subset is selected as the validation set, and the remaining n-1 subsets are used for model training. The trained model is then used as the validation set, and the root mean square error Ri of the predictions for all samples in the validation set is calculated. After looping n times, the average value of the root mean square error of the predictions is calculated.
[0057] Set an error threshold value, if If the error exceeds the preset error threshold, adjust the sample data, check the data attribute calculation results, and iterate again; if... If the error is less than the preset error threshold, the model construction is complete.
[0058] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above method.
[0059] The present invention also provides a computer storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.
[0060] The beneficial effects of this invention are as follows: By calculating the contribution of multiple attributes to the score and the degree of association between attributes, evaluation attributes are screened, resulting in less influence from subjective factors, faster model convergence, and higher accuracy. The use of neural networks to establish a multi-attribute evaluation model avoids the controversy surrounding the subjective setting of thresholds for single attributes in existing technologies. Mathematical statistical methods are used to screen attributes that contribute significantly to the quality of seismic data and to remove attributes that interact with other attributes, improving the scientific rigor of the multi-attribute evaluation. A comprehensive score is obtained by combining manual rating with objective scoring results, achieving quantitative evaluation based on multiple attributes. By calculating the variance inflation factor to quantify the linear correlation between attributes, evaluation attributes without mutual influence are screened, ensuring that evaluation attributes are independent and preventing the model from getting trapped in local optima and generating prediction bias. Attached Figure Description
[0061] Figure 1 This is a diagram illustrating the method steps of the present invention.
[0062] Figure 2 This is a schematic diagram of earthquake data at different levels in Embodiment 2 of the present invention.
[0063] Figure 3 This is a schematic diagram showing the contribution distribution of each feature attribute of work area A to the comprehensive score in Embodiment 2 of the present invention.
[0064] Figure 4 This is a schematic diagram of the prediction effect of the multi-attribute comprehensive evaluation model for work area A in Embodiment 2 of the present invention.
[0065] Figure 5 This is a schematic diagram showing the contribution distribution of each feature attribute of work area B to the comprehensive score in Embodiment 2 of the present invention.
[0066] Figure 6 This is a schematic diagram of the prediction effect of the multi-attribute comprehensive evaluation model for work area B in Embodiment 2 of the present invention. Detailed Implementation
[0067] To clearly illustrate the technical features of this solution, the following detailed implementation method will be used to explain the solution.
[0068] Example 1:
[0069] See Figure 1 As shown in the figure, this embodiment provides an automatic multi-attribute evaluation method for seismic acquisition data, including the following steps:
[0070] S1. Collect seismic data and preprocess it to establish a sample seismic database.
[0071] Preprocessing includes outlier removal, noise reduction, and standardization.
[0072] It should be noted that outlier data refers to values that are significantly inconsistent with the majority of the data. For example, in seismic data, outliers may be caused by equipment malfunctions, acquisition errors, external interference, etc. Statistical methods, such as the standard deviation method or box plot method, are used to identify data points that exceed the normal range and remove them, thereby reducing noise and improving data quality.
[0073] Denoising processes eliminate irrelevant or interfering components in seismic data, making the data more accurate and thus improving the accuracy of subsequent analyses.
[0074] Z-score standardization or min-max standardization is used to unify data of different dimensions and scales, preventing some data from having a disproportionate impact on subsequent scoring due to their large scale, and avoiding the impact of scale differences on the final model or scoring results.
[0075] The sample earthquake database covers more than 95% of the recorded features within the work area, including seismic wave correlation features, temporal features, spatial features, and features related to acquisition conditions.
[0076] It should be noted that seismic wave characteristics are the most fundamental seismic data, including the propagation mode, amplitude, frequency, and velocity of seismic waves. Temporal characteristics mainly involve information such as the time of occurrence of the seismic event, the energy released from the hypocenter, the location of the hypocenter, and the time interval between other seismic events. Spatial characteristics refer to the spatial distribution of seismic activity, including the epicenter location, focal depth, and fault distribution. Acquisition condition-related characteristics refer to factors related to the data acquisition process, including the accuracy of the acquisition equipment, environmental conditions (such as temperature and humidity), and sensor performance.
[0077] S2. Select the characteristic attributes of the sample seismic data and manually score the sample seismic data.
[0078] Characteristic attributes include resolution, gun energy, signal-to-noise ratio, acoustic energy, surface wave energy, industrial interference channel, non-operating channel, and main frequency;
[0079] The specific scoring of the sample seismic data is as follows:
[0080] The sample seismic data were manually rated into four levels: unqualified, qualified, good, and excellent.
[0081] Select the time window where the main target layer is located, calculate the signal-to-noise ratio and resolution of the sample seismic data, normalize them respectively, and take the average value as the objective scoring result;
[0082] Based on the manual rating results of the sample earthquake data, the objective scoring results are segmented and normalized.
[0083] The sample earthquake data are calculated into a comprehensive score based on both manual rating results and objective scoring results, as shown in the following formula:
[0084]
[0085] In the formula, a i The value is the segmented normalized value of the objective score result for sample i.
[0086] Seismic data of different levels, such as Figure 2 As shown.
[0087] It should be noted that segmented normalization of objective scoring results refers to mapping the continuous values of objective scores to specific intervals corresponding to the manual grading based on the four manually divided levels (unqualified, qualified, good, and excellent), so that the automatic scoring is aligned with the manual evaluation standards and the distribution differences of automatic scores within different levels are eliminated.
[0088] S3. Divide the feature attributes into positive and negative attributes according to their impact on the seismic data scoring, and then perform normalization processing.
[0089] Positive attributes are positively correlated with earthquake data scores, while negative attributes are negatively correlated with earthquake data scores.
[0090] Positive attribute normalization uses the maximum value normalization method, expressed by the following formula:
[0091]
[0092] In the formula, x i For the i-th positive attribute value, Let x be the i-th normalized positive attribute value. max This is the maximum value of the attribute.
[0093] The normalization process for negative attributes is as follows: Negative attribute values are sorted in ascending order, and the average of the bottom 10% of attribute values is selected as the standard value for the negative attribute. The opposite of this standard value is then used for normalization. The formula is as follows:
[0094]
[0095] In the formula, y norm The standard value for negative attributes, y i Let N be the i-th negative attribute value, and N be the total number of samples of the last 10% of attribute values after the negative attribute values are sorted in ascending order.
[0096]
[0097] In the formula, Let be the i-th normalized negative attribute value.
[0098] It should be noted that by distinguishing between positive and negative attributes and normalizing them separately, all attributes are defined and evaluated on the same scale, reducing the order-of-magnitude difference between the values of each evaluation attribute, which helps with model fitting.
[0099] S4. Calculate the contribution of feature attributes to the overall score and obtain sensitive attributes;
[0100] Grey relational analysis was performed on the characteristic attributes and the comprehensive score, and the characteristic attributes with a degree of 0.7 or greater were selected as sensitive attributes.
[0101] It should be noted that by using grey relational analysis to exclude weakly correlated attributes (feature attributes with a correlation degree of less than 0.7 with the comprehensive score), the model input dimensions are reduced, and key indicators are focused on, effectively improving the model training efficiency.
[0102] S5. Calculate the degree of association between sensitive attributes, analyze whether there is mutual influence between attributes, screen out evaluation attributes, and build and train a multi-attribute comprehensive evaluation model based on the evaluation attributes.
[0103] The degree of association between each sensitive attribute is obtained by calculating the variance inflation factor (VIF).
[0104]
[0105] In the formula, VIF i Let be the variance inflation factor for the i-th evaluation attribute, where i is the number of evaluation attributes. The correlation coefficient between the i-th evaluation attribute and the other k-1 evaluation attributes is used to treat the i-th evaluation attribute as the explained attribute.
[0106]
[0107] In the formula, The average value of the attribute being explained. Let z be the estimated value of the i-th explained attribute. i Let be the attribute value of the i-th interpreted attribute.
[0108] The judgment criteria are as follows:
[0109] When 1≤VIF≤5, there is no mutual influence between attributes;
[0110] When 5 < VIF ≤ 10, there is a weak mutual influence between attributes;
[0111] When 10 < VIF ≤ 100, there is a strong mutual influence between attributes;
[0112] When VIF > 100, there are serious mutual influences between attributes;
[0113] Sensitive attributes that do not have mutual influence are retained as evaluation attributes.
[0114] It should be noted that mutual interference between attributes leads to large fluctuations in model coefficients, making it difficult to distinguish the true contribution of each attribute and resulting in distorted evaluation results (e.g., the contribution of a certain attribute is incorrectly overestimated or underestimated). By calculating the variance inflation factor, the linear correlation between attributes is quantified, and evaluation attributes that do not influence each other are screened out. This ensures that the evaluation attributes are independent of each other, avoids the model getting trapped in local optima, and prevents prediction bias.
[0115] Using the evaluation attributes selected from the sample earthquake data as input and the comprehensive score as output, the original sample dataset is divided into n subsets and trained into a neural network through n iterations.
[0116] For the i-th iteration, the i-th subset is selected as the validation set, and the remaining n-1 subsets are used for model training. The trained model is then used as the validation set, and the root mean square error Ri of the predictions for all samples in the validation set is calculated. After looping n times, the average value of the root mean square error of the predictions is calculated.
[0117] Set an error threshold value, if If the error exceeds the preset error threshold, adjust the sample data, check the data attribute calculation results, and iterate again; if... If the error is less than the preset error threshold, the model construction is complete.
[0118] It should be noted that this embodiment uses a 3-layer MLP neural network structure (number of input layer nodes = number of evaluation attributes, number of hidden layer nodes = number of input layer nodes × 2, and 1 node in the output layer), the activation function is ReLU (to handle non-linear relationships), and the optimizer is Adam (adaptive learning rate).
[0119] S6. Input the standardized evaluation attribute values and use the multi-attribute comprehensive evaluation model to score them.
[0120] Example 2:
[0121] This embodiment provides an automatic multi-attribute evaluation method for seismic acquisition data. Taking the training of a multi-attribute comprehensive evaluation model for work area A and work area B as an example, the aim is to verify the accuracy of this method in seismic data quality control, including the following steps:
[0122] S1. Collect seismic data and preprocess it to establish a sample seismic database.
[0123] S2. Select the characteristic attributes of the sample seismic data and manually score the sample seismic data. The characteristic attributes include resolution, shot energy, signal-to-noise ratio, acoustic energy, surface wave energy, industrial interference channels, and non-working channels.
[0124] S3. Divide the feature attributes into positive and negative attributes according to their impact on the seismic data scoring, and then perform normalization processing.
[0125] Positive attributes are positively correlated with earthquake data scores, while negative attributes are negatively correlated with earthquake data scores.
[0126] Positive attribute normalization uses the maximum value normalization method, expressed by the following formula:
[0127]
[0128] In the formula, x i For the i-th positive attribute value, Let x be the i-th normalized positive attribute value. max This is the maximum value of the attribute.
[0129] The normalization process for negative attributes is as follows: Negative attribute values are sorted in ascending order, and the average of the bottom 10% of attribute values is selected as the standard value for the negative attribute. The opposite of this standard value is then used for normalization. The formula is as follows:
[0130]
[0131] In the formula, y norm The standard value for negative attributes, y i Let N be the i-th negative attribute value, and N be the total number of samples of the last 10% of attribute values after the negative attribute values are sorted in ascending order.
[0132]
[0133] In the formula, Let be the i-th normalized negative attribute value.
[0134] The positive and negative attribute values and score values of the seismic records in Section A of the work area before and after normalization are shown in Tables 1 and 2, respectively.
[0135] Table 1. Positive and negative attributes and scores of work area A before standardization.
[0136]
[0137]
[0138] Table 2. Standardized Positive and Negative Attributes and Scoring Values for Work Area A
[0139]
[0140] As can be seen from the data in Tables 1 and 2, the order of magnitude difference between the attribute values after normalization is greatly reduced, which is beneficial to the model fitting.
[0141] S4. Calculate the contribution of feature attributes to the overall score and obtain sensitive attributes;
[0142] Grey relational analysis was performed on the characteristic attributes and the comprehensive score, and the characteristic attributes with a degree of 0.7 or greater were selected as sensitive attributes.
[0143] In this embodiment, the contribution of each characteristic attribute of work area A to the overall score is as follows: Figure 3 As shown, the attributes of resolution, gun energy, acoustic energy, surface wave energy, signal-to-noise ratio, and main frequency have a contribution of more than 0.7 and can be used as sensitive attributes; while the contribution of industrial interference channels and non-working channels is low and can be ignored.
[0144] The contribution of each characteristic attribute of work area B to the overall score is as follows: Figure 4 As shown, the attributes of resolution shot energy, surface wave energy, signal-to-noise ratio, non-operating channel, and main frequency have a contribution of more than 0.7 and can be used as sensitive attributes; while the contribution of acoustic wave energy and industrial interference channel is low and can be ignored.
[0145] S5. Calculate the degree of association between sensitive attributes, analyze whether there is mutual influence between attributes, screen out evaluation attributes, and build and train a multi-attribute comprehensive evaluation model based on the evaluation attributes.
[0146] The degree of association between each sensitive attribute is obtained by calculating the variance inflation factor (VIF).
[0147]
[0148] In the formula, VIF i Let be the variance inflation factor for the i-th evaluation attribute, where i is the number of evaluation attributes. The correlation coefficient between the i-th evaluation attribute and the other k-1 evaluation attributes is used to treat the i-th evaluation attribute as the explained attribute.
[0149]
[0150] In the formula, The average value of the attribute being explained. Let z be the estimated value of the i-th explained attribute. i Let be the attribute value of the i-th interpreted attribute. The judgment criteria are as follows:
[0151] When 1≤VIF≤5, there is no mutual influence between attributes;
[0152] When 5 < VIF ≤ 10, there is a weak mutual influence between attributes;
[0153] When 10 < VIF ≤ 100, there is a strong mutual influence between attributes;
[0154] When VIF > 100, there are serious mutual influences between attributes;
[0155] Sensitive attributes that do not have mutual influence are retained as evaluation attributes.
[0156] In this embodiment, the degree of association of each sensitive attribute in work area A is shown in Table 3:
[0157] Table 3. The degree of co-occurrence of each sensitive attribute in work area A
[0158]
[0159] As shown in Table 3, the degree of accompaniment of surface wave energy is greater than 5, indicating that there is an interaction between surface wave energy and other sensitive attributes. This interaction is removed in the evaluation. For work area A, resolution, shot energy, acoustic energy and signal-to-noise ratio are selected as evaluation attributes.
[0160] The degree of association of each sensitive attribute in work area B is shown in Table 4:
[0161] Table 4. The degree of co-occurrence of each sensitive attribute in work area B
[0162]
[0163] As shown in Table 4, the degree of association of all sensitive attributes is less than 5, and they are all retained as evaluation attributes.
[0164] A multi-attribute comprehensive evaluation model was constructed and trained based on the evaluation attributes. The prediction results of the neural network model before and after attribute optimization in work area A and work area B were compared, as shown below. Figure 5 , 6 As shown.
[0165] from Figure 5 , 6 The results show that after attribute filtering, the model's prediction results are significantly improved, and the filtered attributes are closer to the true values. Therefore, it can be concluded that this method can effectively improve the prediction accuracy of multi-attribute comprehensive evaluation models.
[0166] S6. Input the standardized evaluation attribute values and use the multi-attribute comprehensive evaluation model to score them.
[0167] Example 3:
[0168] This is the third embodiment of the present invention, which differs from the first two embodiments in that:
[0169] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art or the current technical solution, can be embodied in the form of a software product. This current computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0170] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0171] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.
[0172] The technical features of this invention not described can be implemented by or using existing technology, and will not be repeated here. Of course, the above description is not a limitation of this invention, and this invention is not limited to the examples above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of this invention should also be within the protection scope of this invention.
Claims
1. A method for automatic evaluation of multiple attributes of seismic acquisition data, characterized in that, Includes the following steps: Collect and preprocess earthquake data to establish a sample earthquake database; The sample seismic data is scored by selecting characteristic attributes of the sample seismic data; The feature attributes are divided into positive and negative attributes according to their impact on the seismic data scoring, and then normalized. Calculate the contribution of feature attributes to the overall score and obtain sensitive attributes; Calculate the degree of association between sensitive attributes, analyze whether there is mutual influence between attributes, screen out evaluation attributes, and build and train a multi-attribute comprehensive evaluation model based on the evaluation attributes. Input the normalized evaluation attribute values and use a multi-attribute comprehensive evaluation model to score them.
2. The automatic multi-attribute evaluation method for seismic acquisition data according to claim 1, characterized in that, The preprocessing includes outlier removal and noise reduction; the earthquake data in the sample earthquake database covers more than 95% of the recorded features in the work area, including seismic wave correlation features, temporal features, spatial features, and acquisition condition-related features.
3. The automatic multi-attribute evaluation method for seismic acquisition data according to claim 1, characterized in that, The characteristic attributes include resolution, gun energy, signal-to-noise ratio, acoustic energy, surface wave energy, industrial interference channel, non-operating channel, and main frequency; The specific steps for scoring the sample seismic data are as follows: The sample seismic data were manually rated into four levels: unqualified, qualified, good, and excellent. Select the time window where the main target layer is located, calculate the signal-to-noise ratio and resolution of the sample seismic data, normalize them respectively, and take the average value as the objective scoring result; Based on the manual rating results of the sample earthquake data, the objective rating results are segmented and normalized. The sample earthquake data are calculated into a comprehensive score based on both manual rating results and objective scoring results, as shown in the following formula: In the formula, a i The value is the segmented normalized value of the objective score result for sample i.
4. The automatic multi-attribute evaluation method for seismic acquisition data according to claim 1, characterized in that, The positive attribute is positively correlated with the earthquake data score, and the negative attribute is negatively correlated with the earthquake data score; The positive attribute normalization process uses the maximum value normalization method. The negative attribute normalization process is as follows: the negative attribute values are arranged in ascending order, the average of the bottom 10% of attribute values is selected as the standard value of the negative attribute, and the negative of the standard value is used for normalization. The formula is expressed as follows: In the formula, y norm The standard value for negative attributes, y i Let N be the i-th negative attribute value, and N be the total number of samples of the last 10% of attribute values after the negative attribute values are sorted in ascending order. In the formula, Let be the i-th normalized negative attribute value.
5. The automatic multi-attribute evaluation method for seismic acquisition data according to claim 1, characterized in that, The calculation of the contribution of feature attributes to the comprehensive score, and the acquisition of sensitive attributes, specifically involves: Grey relational analysis was performed on the characteristic attributes and the comprehensive score, and the characteristic attributes with a degree of 0.7 or greater were selected as sensitive attributes.
6. The automatic multi-attribute evaluation method for seismic acquisition data according to claim 1, characterized in that, The calculation of the degree of association between sensitive attributes, the analysis of whether there is mutual influence between attributes, and the selection of evaluation attributes are specifically as follows: The degree of association between each sensitive attribute is obtained by calculating the variance inflation factor (VIF), and the judgment criteria are as follows: When 1≤VIF≤5, there is no mutual influence between attributes; When 5 < VIF ≤ 10, there is a weak mutual influence between attributes; When 10 < VIF ≤ 100, there is a strong mutual influence between attributes; When VIF > 100, there are serious mutual influences between attributes; Sensitive attributes that do not have mutual influence are retained as evaluation attributes.
7. The automatic multi-attribute evaluation method for seismic acquisition data according to claim 1, characterized in that, The specific steps for constructing and training a multi-attribute comprehensive evaluation model based on evaluation attributes are as follows: Using the evaluation attributes selected from the sample earthquake data as input and the comprehensive score as output, the original sample dataset is divided into n subsets and trained into a neural network through n iterations. For the i-th iteration, the i-th subset is selected as the validation set, and the remaining n-1 subsets are used for model training. The trained model is then used as the validation set, and the root mean square error Ri of the predictions for all samples in the validation set is calculated. After looping n times, the average value of the root mean square error of the predictions is calculated. Set an error threshold value, if If the error exceeds the preset error threshold, adjust the sample data, check the data attribute calculation results, and iterate again; if... If the error is less than the preset error threshold, the model construction is complete.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 7.
9. A computer storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1 to 7.